AutoBalance: An Automatic Balancing Framework for Training Physics-Informed Neural Networks

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: An, Kang, Si, Chenhao, Yan, Ming, Ma, Shiqian
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916996061331456
author An, Kang
Si, Chenhao
Yan, Ming
Ma, Shiqian
author_facet An, Kang
Si, Chenhao
Yan, Ming
Ma, Shiqian
contents Physics-Informed Neural Networks (PINNs) provide a powerful and general framework for solving Partial Differential Equations (PDEs) by embedding physical laws into loss functions. However, training PINNs is notoriously difficult due to the need to balance multiple loss terms, such as PDE residuals and boundary conditions, which often have conflicting objectives and vastly different curvatures. Existing methods address this issue by manipulating gradients before optimization (a "pre-combine" strategy). We argue that this approach is fundamentally limited, as forcing a single optimizer to process gradients from spectrally heterogeneous loss landscapes disrupts its internal preconditioning. In this work, we introduce AutoBalance, a novel "post-combine" training paradigm. AutoBalance assigns an independent adaptive optimizer to each loss component and aggregates the resulting preconditioned updates afterwards. Extensive experiments on challenging PDE benchmarks show that AutoBalance consistently outperforms existing frameworks, achieving significant reductions in solution error, as measured by both the MSE and $L^{\infty}$ norms. Moreover, AutoBalance is orthogonal to and complementary with other popular PINN methodologies, amplifying their effectiveness on demanding benchmarks.
format Preprint
id arxiv_https___arxiv_org_abs_2510_06684
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AutoBalance: An Automatic Balancing Framework for Training Physics-Informed Neural Networks
An, Kang
Si, Chenhao
Yan, Ming
Ma, Shiqian
Machine Learning
Numerical Analysis
Optimization and Control
Physics-Informed Neural Networks (PINNs) provide a powerful and general framework for solving Partial Differential Equations (PDEs) by embedding physical laws into loss functions. However, training PINNs is notoriously difficult due to the need to balance multiple loss terms, such as PDE residuals and boundary conditions, which often have conflicting objectives and vastly different curvatures. Existing methods address this issue by manipulating gradients before optimization (a "pre-combine" strategy). We argue that this approach is fundamentally limited, as forcing a single optimizer to process gradients from spectrally heterogeneous loss landscapes disrupts its internal preconditioning. In this work, we introduce AutoBalance, a novel "post-combine" training paradigm. AutoBalance assigns an independent adaptive optimizer to each loss component and aggregates the resulting preconditioned updates afterwards. Extensive experiments on challenging PDE benchmarks show that AutoBalance consistently outperforms existing frameworks, achieving significant reductions in solution error, as measured by both the MSE and $L^{\infty}$ norms. Moreover, AutoBalance is orthogonal to and complementary with other popular PINN methodologies, amplifying their effectiveness on demanding benchmarks.
title AutoBalance: An Automatic Balancing Framework for Training Physics-Informed Neural Networks
topic Machine Learning
Numerical Analysis
Optimization and Control
url https://arxiv.org/abs/2510.06684